An address is language. A map needs coordinates.
“83 Punggol Central” is meaningful to people because it describes a place through a street-address system. Geographic software needs a latitude and longitude, projected coordinate, parcel, building or network position. The conversion between those worlds is called geocoding.
A geocoded point is not simply found. It is matched.
Quick Read: The Geocoding Mechanism
TEXT LOCATION → STANDARDISATION → CANDIDATE MATCHES → SCORE / RULE → COORDINATE OR FEATURE → MATCH QUALITY
1. Addresses Must Be Parsed
House number, street name, unit, postal code, locality and country can be separated into components. Misspellings, abbreviations and inconsistent formatting complicate matching.
2. Candidate Matches Are Ranked
A geocoder may find several plausible locations. Match scores and rules help select the best candidate, but high score is not the same as guaranteed correctness.
3. Match Type Matters
A result can refer to a rooftop, building entrance, parcel centroid, street segment, postal-code centroid or locality centroid. These locations can differ substantially.
4. Interpolated Street Addresses Are Estimates
Some systems estimate an address position along a street segment using address ranges. The coordinate may be plausible without representing the actual building entrance.
5. Reverse Geocoding Runs the Process Backward
Given coordinates, reverse geocoding returns a nearby address or place label. The returned text is an interpretation of location, not necessarily the exact place the user intended.
6. Geocoding Is Not Positional Accuracy
A precisely formatted coordinate can still come from a weak match. Spatial Uncertainty owns the confidence problem; geocoding owns the conversion from description to geographic position.
7. Postal Codes Can Be Powerful but Coarse
Postal codes can narrow location strongly, but depending on the system they may identify an address, group of addresses or area. Analysts should know what geographic object the code represents.
8. Place Names Can Be Ambiguous
“Springfield,” “Central,” or “Victoria” can refer to many places. Country, region and surrounding context help disambiguate.
9. Historical Addresses Can Break Modern Geocoders
Street names change, buildings disappear and administrative units are redrawn. Historical geography may require period-specific gazetteers rather than modern address databases.
10. Primary Geography: Find the Pin
Children can compare an address written in words with the corresponding point on a map and learn that one representation must be translated into another.
11. Secondary Geography: Compare Match Quality
Students can examine rooftop, street-centre and postal-centroid matches and ask which is adequate for a city-level study versus a building-level study.
12. Advanced Geography: Geocoding Error Can Bias Analysis
If unmatched addresses are concentrated among older buildings, informal settlements or particular populations, missing geocodes can produce systematic rather than random bias.
13. Singapore Example
Singapore’s structured street and postal system can support precise location matching, but analysts should still distinguish between a building, postal location, entrance and parcel when the spatial question is fine-grained.
14. Public Health Example
Patient addresses may be geocoded for exposure studies. Positional error matters if the analysis assigns people to very small environmental zones or distance bands.
15. Business Example
Customer addresses can be mapped to study service areas and market reach. Poor matches can distort catchments, especially when records are incomplete or inconsistent.
16. Hostile Test: “The Software Returned a Coordinate”
What match type? What score? Was there more than one candidate? Does the coordinate represent the building or merely the street? The coordinate is the start of verification, not the end.
17. Where Geocoding Reasoning Breaks
- Coordinate worship: assuming a returned point is exact.
- Match-score absolutism: treating high score as certainty.
- Centroid blindness: failing to distinguish building from area centre.
- Ambiguity blindness: ignoring duplicate place names.
- Historical presentism: using modern gazetteers for old addresses without checking change.
- Missingness blindness: ignoring who or what failed to geocode.
18. Ten Questions for Geocoding
- What text location is being matched?
- How is it standardised?
- What candidate sources are used?
- What match type is returned?
- What is the score or confidence?
- Could another place share the name?
- Is the point a rooftop, centroid or interpolation?
- What records failed to match?
- Does positional error matter at this scale?
- How can uncertain matches be verified?
19. Where This Fits
Spatial Data owns geographic data broadly. Spatial Uncertainty owns confidence limits. This article owns conversion of textual place descriptions into geographic positions.
The Idea to Keep
A pin on a map is trustworthy only when we know what the geocoder actually matched.